You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi! I had a brief look at profiling our code to see what takes the most time in the build process. One of the things we do in that process is building OpenAPI for a bunch of services. That's actually the longest part of the process, at the moment.
One thing stood out to me - there are two removeShapes calls in the call tree, which together take ~54% of the OpenAPI conversion - 324 seconds for the whole thing (we parallelize this by service, but that only gets us so far).
I wonder if there are some optimization opportunities in there. Here's a breakdown ofremoveShapes:
I wonder if there's something obvious we could do about CleanStructureAndUnionMembers.
The text was updated successfully, but these errors were encountered:
Now that I think of it... as your model grows and you have more services, computing scoped models by removing shapes is a hopeless cause - it'll just get consistently slower as more shapes are added.
Shouldn't the scopeModelToService be changed to use a builder that only adds the shapes the service still refers to? That seems much easier than trying to optimize here.
Hi! I had a brief look at profiling our code to see what takes the most time in the build process. One of the things we do in that process is building OpenAPI for a bunch of services. That's actually the longest part of the process, at the moment.
One thing stood out to me - there are two
removeShapes
calls in the call tree, which together take ~54% of the OpenAPI conversion - 324 seconds for the whole thing (we parallelize this by service, but that only gets us so far).I wonder if there are some optimization opportunities in there. Here's a breakdown of
removeShapes
:I wonder if there's something obvious we could do about
CleanStructureAndUnionMembers
.The text was updated successfully, but these errors were encountered: